Simultaneous Matrix Multiplication by Tensor Contraction Aggregation

DOI10.48550/arXiv.2610.26563MaRDI QIDQ7400268arXivFDO

Authors Edgar Solomonik, Markus Bläser

Publication date 19 October 2026

Published in arXiv

Tldr We interpret fast matrix multiplication algorithms as aggregations of tensor contractions and provide new explicit constructions that improve upon Strassen's algorithm and have practical value, improving bounds on rank and border rank of matrix products and quantifying their accuracy in finite precision.

Full work available at URL https://doi.org/10.48550/arXiv.2610.26563

arXiv classification

math.NA

Cites work

• \(0(n^{2.7799})\) complexity for \(n\times n\) approximate matrix multiplication
• A noncommutative algorithm for multiplying 5 X 5 matrices using 103 multiplications
• A Set of Batched Basic Linear Algebra Subprograms and LAPACK Routines
• Approximate Solutions for the Bilinear Form Computational Problem
• Asymptotic tensor rank of graph tensors: beyond matrix multiplication
• Counterexamples to Strassen's direct sum conjecture
• Degeneration and complexity of bilinear maps: Some asymptotic spectra.
• Derivation and analysis of fast bilinear algorithms for convolution
• Fast bilinear algorithms for symmetric tensor contractions
• Fast matrix multiplication and its algebraic neighbourhood
• Fast matrix multiplication is stable
• Gaussian elimination is not optimal
• Improving the numerical stability of fast matrix multiplication
• Matrix multiplication via arithmetic progressions
• Numerical stability and tensor nuclear norm
• On Sets of Integers Which Contain No Three Terms in Arithmetical Progression
• On the algorithmic complexity of associative algebras
• Partial and Total Matrix Multiplication
• Rank and border rank of Kronecker powers of tensors and Strassen's laser method
• Stability of fast algorithms for matrix multiplication
• Tensor rank of the direct sum of two copies of $2 \times 2$ matrix multiplication tensor is 14
• Tensor surgery and tensor rank
• The bilinear complexity and practical algorithms for matrix multiplication
• The Helmert Matrices
• The tensor Rank of \(5 \times 5\) matrices multiplication is bounded by 98 andits border Rank by 89
• Towards automated generation of fast and accurate algorithms for recursive matrix multiplication
• Trilinear aggregating with implicit canceling for a new acceleration of matrix multiplication

This page was built for publication: Simultaneous Matrix Multiplication by Tensor Contraction Aggregation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7400268)